Ai decision method for hotel intelligent voice work order generation and circulation

Abstract

This invention discloses an AI decision-making method for the generation and circulation of intelligent voice work orders in hotels, belonging to the fields of artificial intelligence and hotel service technology. This invention solves the problems of cumbersome manual entry and information errors in existing work order systems. It accurately determines whether a work order is generated through a set of business rules, and effectively improves the efficiency of hotel service resource allocation by calculating the estimated completion time of the work order. By performing multi-dimensional independent quantitative scoring and weighted calculation of each field of the structured work order according to preset standards, and introducing a waiting time decay factor mechanism, combined with a segmented decay strategy based on the hotel service level agreement, the effective comprehensive priority score of all pending work orders is refreshed and recalculated in real time at preset intervals. The work order queue is dynamically rearranged based on the score and generation time, achieving both standardization and accuracy in work order priority ranking, while preventing low-priority work orders from being indefinitely shelved.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and hotel service technology, specifically to an AI decision-making method for generating and circulating intelligent voice work orders in hotels. Background Technology

[0002] With the intelligent and digital upgrade of the hotel industry, the work order system, as the core carrier of hotel accommodation services, directly affects the quality of hotel services and management level through its processing efficiency, data utilization capabilities, and system compatibility.

[0003] Currently, there are still many technical deficiencies in the hotel industry's work order processing. Traditional work orders mostly rely on manual input. The voice requests of guests or employees need to be manually converted into text and filled out as work orders. The operation is cumbersome and prone to information errors. Furthermore, unformatted voice information cannot be directly recognized and utilized by the system. At the same time, there is a lack of a unified identification coding and identification resolution mechanism, which makes it impossible to achieve reliable traceability of the entire life cycle of work orders.

[0004] Therefore, to meet existing needs, an AI decision-making method for intelligent voice-based work order generation and workflow in hotels is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide an AI decision-making method for the generation and circulation of intelligent voice work orders in hotels. This method accurately determines whether a work order will be generated through a set of business rules, and effectively improves the efficiency of hotel service resource allocation by calculating the estimated completion time of the work order. By performing multi-dimensional independent quantitative scoring and weighted calculation on each field of the structured work order according to preset standards, and introducing a waiting time decay factor mechanism, combined with a segmented decay strategy based on the hotel service level agreement, the method refreshes and recalculates the effective comprehensive priority score of all pending work orders in real time at preset intervals. The work order queue is dynamically rearranged based on the score and generation time, achieving both standardization and accuracy in work order priority ranking, while preventing low-priority work orders from being indefinitely shelved, thus solving the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels includes the following steps:

[0008] The system collects voice input information from guests / employees by deploying smart voice terminals in guest rooms, public areas, and on staff terminals, and simultaneously collects data on hotel room status, staff load, and equipment status to obtain initial text information.

[0009] The initial text information and multi-dimensional context information are simultaneously input into the natural language understanding model, combined with the hotel service agreement for semantic parsing, to identify user intent, extract core elements from the text, and output intent analysis results.

[0010] The structured intent object is input into the AI ​​decision engine, which calls the corresponding business rule set according to the intent type; multi-objective optimization calculation is performed by combining real-time big data, and structured decision results are output.

[0011] Based on the decision results, structured work order fields are automatically generated; and each work order is assigned a unique identifier code to achieve the binding of work orders with multi-source associated data;

[0012] Based on the structured work order fields, combined with the preset business rule set and intent analysis results, a multi-dimensional priority decision model is constructed to dynamically calculate the work order priority and complete the sorting.

[0013] Based on the work order identifier code, the identifier is parsed to obtain the full set of work order related information. Combined with the work order type, priority, service personnel skill tags, real-time work status and business rule set, the work order is dispatched to the corresponding personnel or smart devices and the work order process is executed.

[0014] Furthermore, the structured intent object is input into the decision engine, including the following steps:

[0015] Based on the dedicated business rule set and real-time big data, the core optimization objective and auxiliary optimization objective of the multi-objective optimization calculation model are set in the decision engine;

[0016] For the intent type of the structured intent object, assign dynamic weight coefficients to each optimization objective and adaptively adjust them according to the business rule set and the real-time operation status of the hotel;

[0017] The multi-objective optimization calculation model determines whether a work order needs to be generated based on the work order generation conditions in the business rule set and the intent type and requirement content in the intent analysis results.

[0018] By combining real-time employee location / workload / skills and service robot status / location from real-time big data, a list of candidates who meet the requirements for executors is selected, and the comprehensive matching score of each candidate is calculated.

[0019] Based on the processing time of similar historical work orders, the real-time work status of executors, and the timeliness of material allocation, the estimated completion time of the work order is calculated, and a structured decision result is output, which includes: work order generation judgment result, work order suggested priority, estimated completion time, and best executor / or smart device.

[0020] Furthermore, a multi-dimensional priority decision-making model is constructed to dynamically calculate and sort work orders, including the following steps:

[0021] Input the information of each field of the structured work order into the multi-dimensional priority decision model, and independently quantify and score each evaluation dimension according to the preset quantitative scoring standard to obtain the initial score value of each dimension.

[0022] The initial score of each dimension is weighted and calculated with the corresponding dimension weight to obtain the overall priority score of the work order;

[0023] Based on the overall priority score, work orders are divided into multiple priority levels; the level classification criteria are preset in the business rule set and directly matched with the hotel service processing flow.

[0024] All work orders in the same pending queue are sorted by priority according to their comprehensive priority score from high to low, and if the scores are the same, they are sorted by the time the work orders were generated from earliest to latest. The work order priority, sorting results and unique identifier codes are then bound to the work order associated data and updated.

[0025] Furthermore, constructing a multi-dimensional priority decision-making model to dynamically calculate and sort work orders also includes the following steps:

[0026] Introduce a waiting time decay factor mechanism and define effective priorities;

[0027] Set attenuation factor gradient standards and formulate gradient and segmented attenuation strategies based on the hotel service level agreement;

[0028] Input the valid priority scores of all pending work orders into the sorting engine, and obtain the actual waiting time of the work orders in the pending queue in real time, and match the corresponding decay factor.

[0029] The effective priority of all work orders is recalculated at preset intervals. The basic comprehensive priority score is multiplied by the attenuation factor to obtain the effective comprehensive priority score of all work orders. The effective comprehensive priority score is used as the basis for determining the final priority of the work orders.

[0030] Furthermore, constructing a multi-dimensional priority decision-making model to dynamically calculate and sort work orders also includes the following steps:

[0031] The scores of work orders that have entered the pending queue are updated in real time, and the effective comprehensive priority score of newly generated work orders is compared with the real-time effective comprehensive priority scores of all work orders in the current pending work order queue.

[0032] The entire queue of pending work orders is dynamically reordered according to the principle of ranking from highest to lowest score.

[0033] If there are work orders with the same score, they are sorted a second time according to the principle of work order generation time from earliest to latest to ensure the uniqueness and rationality of the queue sorting;

[0034] After sorting, a queue priority sorting position is generated for each work order and synchronously updated to the hotel's smart work order APP.

[0035] Furthermore, constructing a multi-dimensional priority decision-making model to dynamically calculate and sort work orders also includes the following steps:

[0036] The multi-dimensional priority decision-making model pre-configures hotel service scenario evaluation dimensions, including but not limited to urgency, initiating entity, operational impact, and demand type.

[0037] Based on the hotel's operational needs, quantitative scoring standards are preset for each evaluation dimension, and customized adjustments and dynamic updates are made based on real-time big data.

[0038] Based on the preset quantitative scoring criteria, the input structured chemical single-field information is matched and scored independently in terms of dimensions, and a unique initial score value is generated for each evaluation dimension.

[0039] Preset the weights of each evaluation dimension in the multi-dimensional priority decision-making model, and dynamically adjust the weight coefficients according to the hotel's core service needs and scenarios.

[0040] Furthermore, it also includes the following steps:

[0041] The multi-source data generated from each decision-making service is stored in the big data platform and managed in a unified manner.

[0042] The work order processing node and processing result are reported by voice through the intelligent voice terminal. The natural language understanding model is used to perform semantic recognition and format conversion on the feedback voice and automatically update the work order status.

[0043] Simultaneously store and manage all nodes of the work order data in big data, and use identification coding and identification resolution to query and trace the work order status in real time;

[0044] Regularly use big data analytics to analyze service bottlenecks, employee performance, and equipment failure patterns; use the analysis results to iteratively optimize the business rule set and decision-making model.

[0045] Furthermore, it also includes the following steps:

[0046] The stored work order full node data is preprocessed to clean and remove invalid, missing and duplicate data, unstructured data is formatted and transformed, quantitative data is normalized, and related data in the decision-making process is labeled and bound to generate a training dataset suitable for the decision tree model.

[0047] Data is categorized and labeled according to demand type, service scenario, and decision-making stage to construct a hotel service decision tree model;

[0048] The decision tree model is trained and pruned based on the training dataset to optimize the model's branching rules and node decision thresholds.

[0049] Knowledge extraction is performed based on the trained decision tree model, extracting entities, entity attributes, and relationships between entities in the hotel service scenario from node branches, decision rules, and data associations;

[0050] The extracted entities, entity attributes, and relationships are structured and encapsulated to generate the basic graph data for the hotel service knowledge graph, and each set of data is bound with a corresponding data source identifier and decision confidence level.

[0051] The basic graph data interacts in real time with the natural language understanding model and the AI ​​decision engine to continuously perform incremental training and iterative optimization of the decision tree model.

[0052] Furthermore, the structured intent object is input into the AI ​​decision engine for multi-objective optimization calculations, including:

[0053] Obtain the set of business rules to be invoked, and extract the pre-configured mapping table between intent types and optimization target identifiers from the set of business rules. Each record in the mapping table contains an intent type label, a list of core optimization target identifiers, and a list of auxiliary optimization target identifiers.

[0054] Extract the intent type tags from the structured intent object and match them with the mapping table to obtain the initial core optimization target identifier list and the initial auxiliary optimization target identifier list corresponding to the current intent type;

[0055] Obtain the hotel's real-time operating time period tags and the backlog coefficient of the current pending work order queue, and call the preset optimization target filtering rules in the business rule set to filter the real-time operating time period tags and backlog coefficients to obtain target adjustment instructions;

[0056] Update the initial core optimization target identifier list and the initial auxiliary optimization target identifier list based on the target adjustment instruction to obtain the core optimization target set and auxiliary optimization target set corresponding to the current intent type;

[0057] Based on the hotel's real-time operation period tags, the initial weight coefficients of each optimization objective are read from the preset weight configuration table. At the same time, the urgency level tags are extracted from the structured intent objects, and the urgency level tags and the initiating entity identity tags are input into the pre-trained weight correction model for weight correction analysis to obtain the target cascade correction coefficients of each optimization objective.

[0058] The initial weight coefficients are multiplied by the target cascade correction coefficients to generate the dynamic weight coefficients for each optimization objective;

[0059] Collect real-time status data and historical execution efficiency data of each candidate executor from real-time big data, and determine the expected processing time of each candidate executor under different work order types based on the historical execution efficiency data;

[0060] A multi-objective optimization function is constructed with minimizing the weighted response time as the first optimization direction and maximizing the load balancing of executors as the second optimization direction. The weighted response time is determined by the real-time location distance of each candidate executor, the expected processing time, and the dynamic weight coefficient.

[0061] The multi-objective optimization function is iteratively solved in the candidate executor set to obtain the optimal solution set that satisfies the first optimization direction and the second optimization direction;

[0062] The system selects the target executor with the highest matching degree from the optimal solution set, and outputs the expected completion time of the work order based on the expected processing time of the target executor and the current material allocation status.

[0063] Furthermore, the estimated completion time for the output work order includes:

[0064] Obtain the target executor identifier and intent type label from the structured decision results;

[0065] Retrieve the real-time location coordinates and movement speed corresponding to the target executor's identifier from the real-time big data platform;

[0066] Based on the coordinates of the current work order location obtained from the hotel's electronic map, the movement time of the target executor is calculated.

[0067] Access the historical work order database and filter out a set of completed work orders that match the current intent type tag from the historical work order database;

[0068] Extract the pure processing time of each work order in the set of completed work orders, and calculate the average and standard deviation of the pure processing time;

[0069] Obtain the material preparation coefficient in the current material allocation status, and calculate the overall processing time of the work order based on the material preparation coefficient;

[0070] The estimated completion time of the work order is obtained by adding the time spent by the executor moving to the overall processing time of the work order.

[0071] Compared with the prior art, the beneficial effects of the present invention are:

[0072] 1. By accurately determining whether a work order has been generated through a set of business rules, and combining real-time big data to complete the screening of executor candidates and the calculation of comprehensive matching scores, the system scientifically calculates the expected completion time of the work order based on historical data and real-time resource status, and finally outputs structured decision results. This effectively improves the allocation efficiency of hotel service resources, ensures the dynamic adaptation of service decisions to the actual operating status of the hotel, and lays a scientific decision-making foundation for the subsequent generation, circulation and execution of work orders.

[0073] 2. By performing multi-dimensional independent quantitative scoring and weighted calculation of each field of the structured work order according to preset standards, and introducing a waiting time decay factor mechanism, combined with the hotel service level agreement to formulate a segmented decay strategy, the effective comprehensive priority score of all pending work orders is refreshed and recalculated in real time at preset intervals. The work order queue is dynamically rearranged by combining the score with the generation time, generating a unique sorting position for each work order and synchronizing it to the terminal APP. This achieves standardization, dynamism and accuracy in work order priority sorting, effectively preventing low-priority work orders from being shelved indefinitely, and allowing managers to monitor the work order queue sorting status in real time, greatly improving the overall efficiency of hotel work order processing and the rationality of resource allocation.

[0074] 3. By dynamically adjusting and optimizing objectives and weights, and constructing a multi-objective optimization function based on real-time big data, the system achieves precise matching of service resources and scientific prediction of response time. It can adaptively select the optimal executor based on intent type and real-time operational status, effectively improving the overall efficiency and resource utilization of work order processing. This ensures that service decisions are dynamically adapted to the hotel's real-time operational status, providing a scientific decision-making basis for subsequent work order flow and execution.

[0075] 4. By quantifying the executor's movement time and the overall processing time of the work order, the completion time limit of the work order can be accurately calculated. It takes into account the dynamic changes of the executor's real-time location and movement speed, and combines the statistical characteristics of historical processing data and the current material preparation status, so that the prediction of the completion time limit is more in line with the actual operation status. This provides a reliable time basis for subsequent work order scheduling and service commitments, and effectively improves the accuracy of service response and resource allocation. Attached Figure Description

[0076] Figure 1 This is a flowchart of the AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to the present invention. Detailed Implementation

[0077] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] To address the technical challenges of cumbersome manual entry of work orders, which can lead to information errors, the inability of the system to directly recognize and utilize unformatted voice information, and the lack of a unified identification, coding, and parsing mechanism that hinders reliable traceability throughout the entire work order lifecycle, please refer to [link to relevant documentation]. Figure 1 This embodiment provides the following technical solution:

[0079] The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels includes the following steps:

[0080] Step 1: Collect guest / employee voice input information through smart voice terminals deployed in guest rooms, public areas, and employee terminals. Voice input information includes non-formatted voice content such as inquiries, customer needs, complaints, maintenance requests, work order node reports, and information queries. Simultaneously, collect hotel room status, staff load, and equipment status data. Perform noise reduction, sentence segmentation, and speech-to-text processing on the voice information to obtain initial text information.

[0081] Step 2: Simultaneously input the initial text information and multi-dimensional context information into the natural language understanding model. The multi-dimensional context information includes: requester identity and permissions, real-time room status, real-time status of guest room equipment, dialogue history, geographical location, and time information. Combined with the hotel service agreement, perform semantic parsing to identify user intent, extract core elements from the text, and output intent analysis results. The intent analysis results include: room number, user identity, demand type, urgency level, service node, and device type.

[0082] Step 3: Input the structured intent object into the AI ​​decision engine. Based on the intent type, call the corresponding business rule set; for example: repair request rules, customer need rules, and emergency event rules. Combine real-time big data such as: real-time workload and location of employees in each position, service robot status, spare parts inventory, and historical processing time of similar work orders to perform multi-objective optimization calculations, such as: minimizing response time and balancing employee load, and output structured decision results; for example: whether to generate a work order, such as not generating one for simple queries; the priority and expected completion time of the work order; the best executor or execution team, such as assigning it to the nearest idle employee or scheduling a delivery robot; specifically including the following steps:

[0083] Based on the dedicated business rule set and real-time big data, the core optimization objectives of the multi-objective optimization calculation model are set in the decision engine, such as minimizing service response time and balancing the workload of employees in various positions. The auxiliary optimization objectives are such as improving the utilization rate of smart devices, reducing material allocation costs, and matching the skills and needs of the implementers.

[0084] For the intent type of the structured intent object, a dynamic weight coefficient is assigned to each optimization goal, and adaptive adjustments are made based on the business rule set and the hotel's real-time operational status, such as peak / off-peak and overall employee load rate. For example, when handling emergency event intents, the weight coefficient for minimizing service response time is set to 0.6, and the weight coefficients for other goals are allocated proportionally. When handling regular customer need intents, the weight coefficient for balancing the workload of employees in each position is set to 0.5.

[0085] The multi-objective optimization calculation model determines whether a work order needs to be generated based on the work order generation conditions in the business rule set and the intent type and requirement content in the intent analysis results. For example, simple information query intents are directly determined not to generate a work order and trigger the instant voice feedback process; repair, customer needs, and emergency event intents are determined to generate a work order.

[0086] By combining real-time employee location / workload / skills and service robot status / location from real-time big data, a candidate list of executors that meet the requirements is selected, and the comprehensive matching score of each candidate is calculated. Based on the processing time of similar work orders in history, the real-time work status of executors, and the timeliness of material allocation, the expected completion time of the work order is calculated, and a structured decision result is output, which includes: work order generation judgment result, work order suggested priority, expected completion time, and best executor / or smart device.

[0087] The beneficial effects achieved by the above are as follows: the business rule set accurately determines whether a work order is generated, and real-time big data is used to complete the screening of executor candidates and the calculation of comprehensive matching scores; based on historical data and real-time resource status, the estimated completion time of the work order is scientifically calculated, and finally, structured decision results are output, which effectively improves the allocation efficiency of hotel service resources, ensures the dynamic adaptation of service decisions and the actual operating status of the hotel, and lays a scientific decision-making foundation for the subsequent generation, circulation and execution of work orders.

[0088] Step 4: Based on the decision results, automatically generate structured work order fields; and assign a unique identifier code to each work order to achieve the binding of work orders with multi-source related data; the work order includes structured fields such as: unique identifier code, service type, detailed description, location of occurrence, priority, assigned object, and required completion time.

[0089] Step 5: Based on the structured work order fields, combined with the preset business rule set and intent analysis results, construct a multi-dimensional priority decision model to dynamically calculate work order priorities and complete the sorting; specifically including the following steps:

[0090] The multi-dimensional priority decision-making model pre-configures hotel service scenario evaluation dimensions. These dimensions are set in conjunction with the core service needs during a hotel stay, including but not limited to urgency, initiating entity, operational impact, and demand type. Based on hotel operational needs, each evaluation dimension has a pre-defined quantitative scoring standard, such as a scoring range of 0-100 points, which is customized and dynamically updated based on real-time big data. Examples of quantitative scoring standards include: Urgency dimension: room leaks / power outages (90-100 points), routine equipment repairs (50-60 points), general guest service needs (20-30 points); Initiating entity dimension: hotel platinum members (80-90 points), general guests (50-60 points), hotel staff needs (30-40 points); Operational impact dimension: public area malfunctions (80-90 points), single room non-functional needs (30-40 points), peak season service needs (additional 10-20 points). Demand type dimensions: Safety-related needs (90-100 points), Engineering maintenance needs (50-70 points), Material distribution needs (20-40 points).

[0091] Based on the preset quantitative scoring standards, the input structured work order field information is matched and scored independently, generating a unique initial score value for each evaluation dimension; the scoring results are stored in real time in the work order associated data, supporting score traceability and manual review; the dimension weights of each evaluation dimension of the multi-dimensional priority decision model are preset, with weight values ​​being decimals between 0 and 1, and the sum of all dimension weights being 1. According to the hotel's core service requirements, the weight coefficients are dynamically adjusted according to the scenario. The configuration example is as follows: urgency dimension (0.4), initiating entity dimension (0.2), operational impact dimension (0.3), and demand type dimension (0.1).

[0092] The structured work order's field information is input into a multi-dimensional priority decision model. Each evaluation dimension is independently quantified and scored according to a preset quantitative scoring standard, yielding an initial score for each dimension. The initial scores for each dimension are then weighted with their corresponding dimension weights to obtain the work order's overall priority score. Based on the overall priority score, work orders are divided into multiple priority levels, such as: Level 1 Urgent, Level 2 High Priority, Level 3 Routine, and Level 4 Low Priority. The level classification criteria are preset in the business rule set and directly matched with the hotel service processing flow. All work orders in the same pending queue are sorted by overall priority score from highest to lowest, and for those with the same score, by work order generation time from earliest to latest, completing the work order priority ranking. The work order priority, ranking result, and unique identifier code are then bound again and updated in the work order association data, providing a core basis for subsequent work order flow and execution dispatch.

[0093] A waiting time decay factor mechanism is introduced to define effective priority. The decay factor is a dynamic coefficient greater than 1, and its value is positively correlated with the actual waiting time of the work order. That is, the longer the waiting time of the work order, the larger the decay factor value. The preset base decay factor is 1.0. The waiting time is calculated from the time the work order is generated and enters the pending queue, and is accumulated in minutes. A decay factor gradient standard is set, and a gradient and segmented decay strategy is formulated according to the hotel service level agreement. For example: the first stage: waiting 0-30 minutes (decay factor 1.0, linear and slow increase to prevent new work orders from being over-prioritized), the second stage: waiting 30-120 minutes (decay factor 1.2, exponential acceleration to ensure timely processing), and the third stage: waiting more than 120 minutes (decay factor 1.5, enters the emergency processing queue, and automatically triggers management alarms).

[0094] The effective priority scores of all pending work orders are input into the sorting engine, and the actual waiting time of each work order in the pending queue is obtained in real time. The corresponding decay factor is matched. The effective priority of all work orders is recalculated at preset intervals, such as every 5 minutes. The basic comprehensive priority score is multiplied by the decay factor to obtain the effective comprehensive priority score of all work orders. The effective comprehensive priority score is used as the basis for the final priority determination of the work orders. Abnormal queue statuses, such as backlog of a certain type of work orders, are also identified.

[0095] The system refreshes the scores of work orders already in the pending queue in real time, with a configurable refresh frequency of once per minute, to prevent low-priority work orders from being indefinitely shelved due to continuous waiting. The system compares the effective comprehensive priority score of newly generated work orders with the real-time effective comprehensive priority scores of all work orders in the current pending queue. The entire pending work order queue is dynamically reordered according to the score from highest to lowest. If work orders have the same score, they are reordered according to their creation time from earliest to latest, ensuring the uniqueness and rationality of the queue sorting. After sorting, a queue priority ranking position is generated for each work order and simultaneously updated in the hotel's smart work order APP, allowing managers to view the queue sorting results and work order score details in real time.

[0096] Step Six: Based on the work order identifier code, perform identifier parsing to obtain the full set of work order related information. Combined with the work order type, priority, service personnel skill tags, real-time work status, and business rule set, the work order is dispatched to the corresponding personnel or smart devices, and the work order process is executed. The work order achieves real-time synchronous flow across platforms and systems under the microservice architecture, supporting seamless integration with hotel PMS systems, delivery robots, intelligent guest room control systems, and intelligent contract management systems.

[0097] Step 7: Store the multi-source data generated from each decision-making service into a big data platform, such as: original voice, interpreted intent, decision basis, execution log, completion time, and satisfaction level, for unified big data storage and management; report work order processing nodes and feedback results via intelligent voice terminals using voice, and use natural language understanding models to perform semantic recognition and formatted conversion on the feedback voice, automatically updating the work order status; simultaneously store and manage all work order node data in big data, and perform real-time querying and tracing of work order status through identification coding and identification resolution; regularly analyze service bottlenecks, employee performance, and equipment failure patterns through big data analysis; use the analysis results to iteratively optimize the business rule set and decision-making model, achieving system self-evolution and continuous iteration of AI decision-making capabilities, forming an intelligent decision-making system.

[0098] Step 8: Preprocess the stored work order full-node data, cleaning and removing invalid, missing, and duplicate data; formatting and converting unstructured data; normalizing quantitative data; and labeling and binding related data in the decision-making process to generate a training dataset suitable for the decision tree model. The work order full-node data includes: structured / unstructured data collected throughout the work order process; intermediate and result data from the entire AI decision-making process; specifically, initial speech-to-text text, core elements of intent analysis, business rule set call records, multi-objective optimization calculation parameters, priority scoring / weight / decay factor data, work order assignment and execution logs, work order completion time, and satisfaction feedback. Real-time big data on equipment, personnel, and inventory; data is categorized and labeled according to demand type, service scenario, and decision-making stage to construct a hotel service decision tree model. The root node is based on core AI decision-making objectives such as minimizing response time, balancing load, and improving service matching. Intermediate nodes are key decision dimensions for work order processing, such as demand type, urgency, initiating entity, operational impact, resource status, and scenario characteristics. Leaf nodes are decision results, such as work order generation determination, priority level, executor assignment, and completion time calculation. The decision tree model is trained and pruned based on the training dataset to optimize branching rules and node decision thresholds, ensuring that the model's decision logic aligns with actual hotel service scenarios. Adaptability; Based on the trained decision tree model, knowledge extraction is performed to extract entities, entity attributes, and inter-entity relationships from the hotel service scenario from node branches, decision rules, and data associations. Entities include: guests, employees, smart devices, work orders, service types, and device types; entity attributes include: guest level, employee skills, work order priority, and device status; inter-entity relationships include: guest initiation - work order type, work order type - matching executor, urgency - priority weight, and resource status - decision result. The extracted entities, entity attributes, and relationships are structured and encapsulated to generate the basic graph data for a hotel service knowledge graph, and each data set is then... By binding corresponding data source identifiers and decision confidence levels, the knowledge graph can be traced and its credibility assessed. Real-time interaction between the basic graph data and natural language understanding models and AI decision engines provides knowledge support for semantic parsing, intent recognition, business rule invocation, and multi-objective optimization calculations. Based on real-time new data from work order processing and dynamic data from the decision-making process, the decision tree model undergoes continuous incremental training and iterative optimization. New entities and relationships are extracted simultaneously, and existing hotel service knowledge graphs are supplemented with nodes, updated with relationships, and have their confidence levels corrected. This enables the knowledge graph to self-iterate and self-improve, ensuring real-time matching between the graph content and actual hotel service operations and AI decision-making logic.

[0099] The beneficial effects achieved by the above are as follows: By performing multi-dimensional independent quantitative scoring and weighted calculation of each field of the structured work order according to preset standards, and introducing a waiting time decay factor mechanism, combined with the hotel service level agreement to formulate a segmented decay strategy, the effective comprehensive priority score of all pending work orders is refreshed and recalculated in real time at preset intervals. The work order queue is dynamically rearranged by combining the score with the generation time, generating a unique sorting position for each work order and synchronizing it to the terminal APP. This not only achieves the standardization, dynamism and accuracy of work order priority sorting, effectively avoiding the indefinite shelving of low-priority work orders, but also allows managers to grasp the sorting status of the work order queue in real time, greatly improving the overall efficiency of hotel work order processing and the rationality of resource allocation.

[0100] Working principle: The system uses a natural language understanding model to semantically analyze real-time hotel service data and outputs structured intent objects. The model then calls upon the corresponding business rule set, combines real-time big data to dynamically assign weights to a multi-objective optimization calculation model, and completes work order generation, executor selection, and completion time calculation, outputting standardized structured decision results. Based on the decision results, structured work order fields are automatically generated and each work order is assigned a unique identifier. A multi-dimensional priority decision model then performs quantitative scoring, weighted calculation, and level classification of the work orders. A waiting time decay factor mechanism is introduced to refresh and dynamically sort work order priorities in real time. Combined with business rules, intelligent work order dispatch and execution are completed. A knowledge graph is constructed based on the work order data, providing knowledge support for natural language understanding and AI decision-making, forming an intelligent closed-loop working system.

[0101] This embodiment provides an AI decision-making method for generating and circulating intelligent voice work orders in hotels. It inputs structured intent objects into the AI ​​decision engine for multi-objective optimization calculations, including:

[0102] Obtain the set of business rules to be invoked, and extract the pre-configured mapping table between intent types and optimization target identifiers from the set of business rules. Each record in the mapping table contains an intent type label, a list of core optimization target identifiers, and a list of auxiliary optimization target identifiers.

[0103] Extract the intent type tags from the structured intent object and match them with the mapping table to obtain the initial core optimization target identifier list and the initial auxiliary optimization target identifier list corresponding to the current intent type;

[0104] Obtain the hotel's real-time operating time period tags and the backlog coefficient of the current pending work order queue, and call the preset optimization target filtering rules in the business rule set to filter the real-time operating time period tags and backlog coefficients to obtain target adjustment instructions;

[0105] Update the initial core optimization target identifier list and the initial auxiliary optimization target identifier list based on the target adjustment instruction to obtain the core optimization target set and auxiliary optimization target set corresponding to the current intent type;

[0106] Based on the hotel's real-time operation period tags, the initial weight coefficients of each optimization objective are read from the preset weight configuration table. At the same time, the urgency level tags are extracted from the structured intent objects, and the urgency level tags and the initiating entity identity tags are input into the pre-trained weight correction model for weight correction analysis to obtain the target cascade correction coefficients of each optimization objective.

[0107] The initial weight coefficients are multiplied by the target cascade correction coefficients to generate the dynamic weight coefficients for each optimization objective;

[0108] Collect real-time status data and historical execution efficiency data of each candidate executor from real-time big data, and determine the expected processing time of each candidate executor under different work order types based on the historical execution efficiency data;

[0109] A multi-objective optimization function is constructed with minimizing the weighted response time as the first optimization direction and maximizing the load balancing of executors as the second optimization direction. The weighted response time is determined by the real-time location distance of each candidate executor, the expected processing time, and the dynamic weight coefficient.

[0110] The multi-objective optimization function is iteratively solved in the candidate executor set to obtain the optimal solution set that satisfies the first optimization direction and the second optimization direction;

[0111] The system selects the target executor with the highest matching degree from the optimal solution set, and outputs the expected completion time of the work order based on the expected processing time of the target executor and the current material allocation status.

[0112] In this embodiment, the urgency level label and the initiating entity identity label are input into the pre-trained weight correction model for weight correction analysis to obtain the target cascade correction coefficients for each optimization objective, including the following steps:

[0113] A pre-configured mapping table of urgency level, initiating entity combination, and correction coefficient is provided. Each row in the mapping table corresponds to a combination of urgency level label and initiating entity identity label, and each column corresponds to an optimization target. The cell value is the correction coefficient of the optimization target under the corresponding combination.

[0114] The urgency level tag is concatenated with the initiator's identity tag to generate a combined query key;

[0115] Perform an exact match lookup in the mapping table based on the combined query key to retrieve the target row data corresponding to the combined query key;

[0116] If the target row data exists, extract all the correction coefficients corresponding to the optimization targets from the target row data, and assemble them into a correction coefficient vector according to the preset order of the optimization targets;

[0117] If the target row data does not exist, the urgent level correction base value corresponding to the urgent level label and the initiator correction base value corresponding to the initiator identity label are obtained respectively. The urgent level correction base value and the initiator correction base value are input into the preset coefficient fusion function, and the weighted sum is independently calculated for each optimization target based on the coefficient fusion function to generate the initial correction coefficient of each optimization target. The initial correction coefficients of all optimization targets are normalized to form a correction coefficient vector.

[0118] The correction coefficient vector is output as the target concatenated correction coefficient.

[0119] In this embodiment, the intent type label refers to the identifier that classifies the user's voice intent, such as categories like repair request, customer need, and inquiry.

[0120] In this embodiment, the core optimization target identifier list refers to the set of identifiers set as the main optimization target in multi-objective optimization, such as minimizing response time.

[0121] In this embodiment, the auxiliary optimization target identifier list refers to the set of identifiers for secondary optimization targets, such as balancing the load of executors.

[0122] In this embodiment, the real-time operating period label refers to an identifier that classifies the hotel's current operating period, such as peak period or off-peak period.

[0123] In this embodiment, the backlog coefficient refers to a quantitative value that reflects the congestion level of the current work order queue, which is calculated from parameters such as the ratio of the number of work orders to the number of executors.

[0124] In this embodiment, the optimization target selection rule refers to the pre-set rules in the business rule set, which are used to dynamically adjust the priority of optimization targets according to the real-time operational status.

[0125] In this embodiment, the target adjustment instruction refers to an instruction generated according to the screening rules that indicates which auxiliary optimization targets need to be adjusted to core optimization targets.

[0126] In this embodiment, the preset weight configuration table refers to a pre-set table that assigns initial weights to each optimization objective according to different operating periods.

[0127] In this embodiment, the urgency level label refers to a label that identifies the urgency level of the work order, such as high, medium, or low.

[0128] In this embodiment, the initiating entity identity tag refers to a tag that identifies the identity category of the work order initiator, such as VIP guest, ordinary guest, employee, etc.

[0129] In this embodiment, the pre-trained weight correction model refers to a model that is pre-trained and used to correct the initial weights according to the urgency and the initiating entity and output cascade correction coefficients.

[0130] In this embodiment, the target cascade correction coefficient refers to the coefficient output by the weight correction model, which is used to adjust the weights of each optimization target.

[0131] In this embodiment, the dynamic weight coefficient refers to the actual weight used for multi-objective optimization obtained by multiplying the initial weight by the cascaded correction coefficient.

[0132] In this embodiment, the candidate executor refers to an employee or smart device that may be assigned to handle work orders. Real-time status data refers to dynamic information such as the candidate executor's current location, load, and working status.

[0133] In this embodiment, historical execution efficiency data refers to statistical data such as the time taken by candidate executors to process similar work orders in the past.

[0134] In this embodiment, the expected processing time refers to the time required for a candidate executor to process the current work order, as predicted based on historical efficiency data.

[0135] In this embodiment, the weighted response time refers to the response time index calculated by comprehensively considering distance, expected processing time, and dynamic weight.

[0136] In this embodiment, the executor load balancing degree refers to an indicator that measures the degree of evenness in task distribution among executors.

[0137] In this embodiment, the multi-objective optimization function refers to a mathematical function constructed by combining multiple optimization objectives, used to solve for the optimal executor.

[0138] In this embodiment, the optimal solution set refers to the set of non-dominated solutions that satisfy multiple optimization directions, obtained by solving a multi-objective optimization function.

[0139] In this embodiment, the target executor refers to the executor with the highest matching degree selected from the optimal solution set.

[0140] In this embodiment, the material allocation status refers to the availability or readiness of the materials required to complete the work order, which may affect the processing time.

[0141] The working principle and beneficial effects of the above technical solution are as follows: By dynamically adjusting and optimizing objectives and weights, and combining real-time big data to construct a multi-objective optimization function, the solution achieves accurate matching of service resources and scientific prediction of response time. It can adaptively select the optimal executor based on intent type and real-time operational status, effectively improving the overall efficiency and resource utilization of work order processing, ensuring that service decisions are dynamically adapted to the hotel's real-time operational status, and providing a scientific decision-making basis for subsequent work order flow and execution.

[0142] This embodiment provides an AI decision-making method for generating and circulating intelligent voice work orders in hotels, outputting the estimated completion time of the work orders, including:

[0143] Obtain the target executor identifier and intent type label from the structured decision results;

[0144] Retrieve the real-time location coordinates and movement speed corresponding to the target executor's identifier from the real-time big data platform;

[0145] Based on the coordinates of the current work order location obtained from the hotel's electronic map, the movement time of the target executor is calculated.

[0146]

[0147] in, Indicates the time taken for the target executor to move; The x-coordinate value representing the location where the current work order occurred; The x-coordinate value representing the real-time location of the target executor; The vertical coordinate value represents the location where the current work order occurred; The vertical coordinate value represents the real-time location of the target executor; Indicates the movement speed of the target executor;

[0148] Access the historical work order database and filter out a set of completed work orders that match the current intent type tag from the historical work order database;

[0149] Extract the pure processing time of each work order in the set of completed work orders, and calculate the average and standard deviation of the pure processing time;

[0150] Obtain the material preparation coefficient in the current material allocation status, and calculate the overall processing time of the work order based on the material preparation coefficient;

[0151] ;

[0152] in, Indicates the overall processing time of the work order; This represents the average pure processing time; The standard deviation of pure processing time; This represents the preset risk tolerance coefficient, and its value range is (0, 3). This represents the material preparation coefficient, and its value range is (0,1].

[0153] The estimated completion time of the work order is obtained by adding the time spent by the executor moving to the overall processing time of the work order.

[0154] In this embodiment, This indicates that the supplies are in place. This indicates that resources need to be allocated, and the smaller the value, the longer the allocation time.

[0155] In this embodiment, the movement time refers to the estimated time required for the target executor to move from its current location to the location where the work order occurred.

[0156] In this embodiment, the pure processing time refers to the actual time spent by the executor from arriving at the site to completing the service, as recorded in the historical work order database.

[0157] In this embodiment, the material preparation coefficient refers to a quantitative value that reflects the availability or preparation status of the materials required to complete the current work order. The value range is (0,1]. A value of 1 indicates that the materials are in place, and the smaller the value, the longer the time required for material allocation.

[0158] In this embodiment, the preset risk tolerance coefficient refers to a preset parameter used to adjust the fluctuation range of historical processing time. The value ranges from 0 to 3 and is used to control the tolerance level for uncertainty in processing time.

[0159] In this embodiment, the comprehensive processing time refers to the estimated processing time of the work order calculated based on historical data, fluctuation tolerance, and material preparation.

[0160] In this embodiment, the estimated completion time refers to the total estimated time from the generation of the work order to its completion, obtained by adding the time spent by the executor to the overall processing time.

[0161] The working principle and beneficial effects of the above technical solution are as follows: By quantifying the executor's movement time and the overall processing time of the work order, the completion time limit of the work order can be accurately calculated. It takes into account the dynamic changes of the executor's real-time location and movement speed, and combines the statistical characteristics of historical processing data and the current material preparation status, so that the prediction of the completion time limit is more in line with the actual operation status, providing a reliable time basis for subsequent work order scheduling and service commitments, and effectively improving the accuracy of service response and resource allocation.

[0162] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or high-voltage switchgear that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or high-voltage switchgear.

[0163] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI decision-making method for generating and circulating intelligent voice work orders in hotels, characterized in that, Includes the following steps: The system collects voice input information from guests / employees by deploying smart voice terminals in guest rooms, public areas, and on staff terminals, and simultaneously collects data on hotel room status, staff load, and equipment status to obtain initial text information. The initial text information and multi-dimensional context information are simultaneously input into the natural language understanding model, combined with the hotel service agreement for semantic parsing, to identify user intent, extract core elements from the text, and output intent analysis results. The structured intent object is input into the AI ​​decision engine, which calls the corresponding business rule set according to the intent type; multi-objective optimization calculation is performed by combining real-time big data, and structured decision results are output. Based on the decision results, automatically generate structured chemical single field; Each work order is assigned a unique identifier code to bind the work order to multi-source associated data; Based on the structured work order fields, combined with the preset business rule set and intent analysis results, a multi-dimensional priority decision model is constructed to dynamically calculate the work order priority and complete the sorting. Based on the work order identifier code, the identifier is parsed to obtain the full set of work order related information. Combined with the work order type, priority, service personnel skill tags, real-time work status and business rule set, the work order is dispatched to the corresponding personnel or smart devices and the work order process is executed.

2. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 1, characterized in that, Inputting structured intent objects into the decision engine includes the following steps: Based on the dedicated business rule set and real-time big data, the core optimization objective and auxiliary optimization objective of the multi-objective optimization calculation model are set in the decision engine; For the intent type of the structured intent object, assign dynamic weight coefficients to each optimization objective and adaptively adjust them according to the business rule set and the real-time operation status of the hotel; The multi-objective optimization calculation model determines whether a work order needs to be generated based on the work order generation conditions in the business rule set and the intent type and requirement content in the intent analysis results. By combining real-time employee location / workload / skills and service robot status / location from real-time big data, a list of candidates who meet the requirements for executors is selected, and the comprehensive matching score of each candidate is calculated. Based on the processing time of similar historical work orders, the real-time work status of executors, and the timeliness of material allocation, the estimated completion time of the work order is calculated, and a structured decision result is output, which includes: work order generation judgment result, work order suggested priority, estimated completion time, and best executor / or smart device.

3. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 1, characterized in that, Construct a multi-dimensional priority decision-making model to dynamically calculate and sort work orders, including the following steps: Input the information of each field of the structured work order into the multi-dimensional priority decision model, and independently quantify and score each evaluation dimension according to the preset quantitative scoring standard to obtain the initial score value of each dimension. The initial score of each dimension is weighted and calculated with the corresponding dimension weight to obtain the overall priority score of the work order; Based on the overall priority score, work orders are divided into multiple priority levels; the level classification criteria are preset in the business rule set and directly matched with the hotel service processing flow. All work orders in the same pending queue are sorted by priority according to their comprehensive priority score from high to low, and if the scores are the same, they are sorted by the time the work orders were generated from earliest to latest. The work order priority, sorting results and unique identifier codes are then bound to the work order associated data and updated.

4. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 3, characterized in that, The construction of a multi-dimensional priority decision-making model, which dynamically calculates and sorts work orders, also includes the following steps: Introduce a waiting time decay factor mechanism and define effective priorities; Set attenuation factor gradient standards and formulate gradient and segmented attenuation strategies based on the hotel service level agreement; Input the valid priority scores of all pending work orders into the sorting engine, and obtain the actual waiting time of the work orders in the pending queue in real time, and match the corresponding decay factor. The effective priority of all work orders is recalculated at preset intervals. The basic comprehensive priority score is multiplied by the attenuation factor to obtain the effective comprehensive priority score of all work orders. The effective comprehensive priority score is used as the basis for determining the final priority of the work orders.

5. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 4, characterized in that, The construction of a multi-dimensional priority decision-making model, which dynamically calculates and sorts work orders, also includes the following steps: The scores of work orders that have entered the pending queue are updated in real time, and the effective comprehensive priority score of newly generated work orders is compared with the real-time effective comprehensive priority scores of all work orders in the current pending work order queue. The entire queue of pending work orders is dynamically reordered according to the principle of ranking from highest to lowest score. If there are work orders with the same score, they are sorted a second time according to the principle of work order generation time from earliest to latest to ensure the uniqueness and rationality of the queue sorting; After sorting, a queue priority sorting position is generated for each work order and synchronously updated to the hotel's smart work order APP.

6. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 5, characterized in that, The construction of a multi-dimensional priority decision-making model, which dynamically calculates and sorts work orders, also includes the following steps: The multi-dimensional priority decision-making model pre-configures hotel service scenario evaluation dimensions, including but not limited to urgency, initiating entity, operational impact, and demand type. Based on the hotel's operational needs, quantitative scoring standards are preset for each evaluation dimension, and customized adjustments and dynamic updates are made based on real-time big data. Based on the preset quantitative scoring criteria, the input structured chemical single-field information is matched and scored independently in terms of dimensions, and a unique initial score value is generated for each evaluation dimension. Preset the weights of each evaluation dimension in the multi-dimensional priority decision-making model, and dynamically adjust the weight coefficients according to the hotel's core service needs and scenarios.

7. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 1, characterized in that, It also includes the following steps: The multi-source data generated from each decision-making service is stored in the big data platform and managed in a unified manner. The work order processing node and processing result are reported by voice through the intelligent voice terminal. The natural language understanding model is used to perform semantic recognition and format conversion on the feedback voice and automatically update the work order status. Simultaneously store and manage all nodes of the work order data in big data, and use identification coding and identification resolution to query and trace the work order status in real time; Regularly use big data analytics to analyze service bottlenecks, employee performance, and equipment failure patterns; use the analysis results to iteratively optimize the business rule set and decision-making model.

8. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 7, characterized in that, It also includes the following steps: The stored work order full node data is preprocessed to clean and remove invalid, missing and duplicate data, unstructured data is formatted and transformed, quantitative data is normalized, and related data in the decision-making process is labeled and bound to generate a training dataset suitable for the decision tree model. Data is categorized and labeled according to demand type, service scenario, and decision-making stage to construct a hotel service decision tree model; The decision tree model is trained and pruned based on the training dataset to optimize the model's branching rules and node decision thresholds. Knowledge extraction is performed based on the trained decision tree model, extracting entities, entity attributes, and relationships between entities in the hotel service scenario from node branches, decision rules, and data associations; The extracted entities, entity attributes, and relationships are structured and encapsulated to generate the basic graph data for the hotel service knowledge graph, and each set of data is bound with a corresponding data source identifier and decision confidence level. The basic graph data interacts in real time with the natural language understanding model and the AI ​​decision engine to continuously perform incremental training and iterative optimization of the decision tree model.

9. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 1, characterized in that, The structured intent object is input into the AI ​​decision engine for multi-objective optimization calculations, including: Obtain the set of business rules to be invoked, and extract the pre-configured mapping table between intent types and optimization target identifiers from the set of business rules. Each record in the mapping table contains an intent type label, a list of core optimization target identifiers, and a list of auxiliary optimization target identifiers. Extract the intent type tags from the structured intent object and match them with the mapping table to obtain the initial core optimization target identifier list and the initial auxiliary optimization target identifier list corresponding to the current intent type; Obtain the hotel's real-time operating time period tags and the backlog coefficient of the current pending work order queue, and call the preset optimization target filtering rules in the business rule set to filter the real-time operating time period tags and backlog coefficients to obtain target adjustment instructions; Update the initial core optimization target identifier list and the initial auxiliary optimization target identifier list based on the target adjustment instruction to obtain the core optimization target set and auxiliary optimization target set corresponding to the current intent type; Based on the hotel's real-time operation period tags, the initial weight coefficients of each optimization objective are read from the preset weight configuration table. At the same time, the urgency level tags are extracted from the structured intent objects, and the urgency level tags and the initiating entity identity tags are input into the pre-trained weight correction model for weight correction analysis to obtain the target cascade correction coefficients of each optimization objective. The initial weight coefficients are multiplied by the target cascade correction coefficients to generate the dynamic weight coefficients for each optimization objective; Collect real-time status data and historical execution efficiency data of each candidate executor from real-time big data, and determine the expected processing time of each candidate executor under different work order types based on the historical execution efficiency data; A multi-objective optimization function is constructed with minimizing the weighted response time as the first optimization direction and maximizing the load balancing of executors as the second optimization direction. The weighted response time is determined by the real-time location distance of each candidate executor, the expected processing time, and the dynamic weight coefficient. The multi-objective optimization function is iteratively solved in the candidate executor set to obtain the optimal solution set that satisfies the first optimization direction and the second optimization direction; The system selects the target executor with the highest matching degree from the optimal solution set, and outputs the expected completion time of the work order based on the expected processing time of the target executor and the current material allocation status.

10. The AI ​​decision-making method for generating and circulating intelligent voice work orders in hotels according to claim 9, characterized in that, The estimated completion time for output work orders includes: Obtain the target executor identifier and intent type label from the structured decision results; Retrieve the real-time location coordinates and movement speed corresponding to the target executor's identifier from the real-time big data platform; Based on the coordinates of the current work order location obtained from the hotel's electronic map, the movement time of the target executor is calculated. Access the historical work order database and filter out a set of completed work orders that match the current intent type tag from the historical work order database; Extract the pure processing time of each work order in the set of completed work orders, and calculate the average and standard deviation of the pure processing time; Obtain the material preparation coefficient in the current material allocation status, and calculate the overall processing time of the work order based on the material preparation coefficient; The estimated completion time of the work order is obtained by adding the time spent by the executor moving to the overall processing time of the work order.

Images